10.7a Catalyst Beta Drivers Re-Thread

[railven]

What the hell why did the 10.7a beta drivers thread get locked? There are countless other threads with more poo flinging and de-railing than that one and they are open for anyone to wade through.

I was hoping to see more Radeon user's opinions of AA on SC2.


I used a 5870 2GB stock settings on a Core i7 920 with 6GBs of DDR3 RAM (sorry if I'm not specific) gaming at 1920x1080.

Girlfriend uses a GTX 460 on an AMD quad core (965) with 6GBs of DDR3 RAM, everything stock, 1650x1080.

I get around 70FPS AA 4x on, she hits gets around 60FPS.

Either way, when all is said and done - I like the AA on and so does she. All the fanboy stuff aside, glad Nvidia and ATI both delivered on this since now we can enjoy it.

 

[bryanW1995]

it got locked because too many people have personal agendas in here.

 

[Skurge]

So I guess I should be getting abut half that? My copy hasn't arrived yet, so I can't test.

 

[Gikaseixas]

Railven can you bench again without AA turned on? I just wanna see how big of a hit do you get on your system. My copy doesn't get here till next week or so.

 

[GaiaHunter]

So I guess I should be getting abut half that? My copy hasn't arrived yet, so I can't test.

I hope you aren't worried - half of that is more than enough for a RTS.

I remember seeing threads where people debated that more than a certain number of fps was actually prejudicial for RTS.

 

[netxzero64]

many of my friends didn't feel any changes from 10.4 to 10.7

 

[Sind]

Messing around tonite with some of the settings with 4x and 8x AA. Without any aa on a skirm map vs easy ai loaded up and went on the attack FPS would be around 80-90 without any action.

As I headed into and cleared the AI base this dipped to as low as 53 fps. I saved and redid with 4x and 8x, 4x never dropped below 38fp, 8x lowest was 23 and seemed almost stuttery at that point. All AA was with Adaptive Multi. The AI didn't really have many units to fight back with so I couldn't really get a feel for two loaded sides going at each other and what the impact would be.

I haven't done any matches or anything, just started playing SC2 tonite basically and was doing some campaigns which was all on 4x and I didn't notice any drops below 48fps. Dunno where people are pulling 4x AA@70fps at, maybe at the start of matches etc, but dips will happen gradually as units fill up etc.

On the bridge for example at 4xAA it hovers around 48fps, with it off its up in the 80s.

Gonna try campaign tomorrow at 8x and see how that is, everything was fine otherwise and super smooth at 4xAA regardless of 40s, or 50s.

Seems like enabling AA to 4x you take a almost 45-50percent; hit in performance and to 8x you will lose another 10-15fps from my short little time messing around. I was only looking for minimums and if there was a stuttering point, 80+fps means nothing in a RTS imo, its how low can you go with all eye candy on without stuttering in huge battles is how I configure my RTS, shooters/mmos I do differently.

All in all I really don't know wtf Blizzard was thinking graphically (well I do, its all about $$$) but they really could have invested more time getting things more optimised it seems and work closer with ATI and Nvidia, the whole frying of cards of those whom are tech illiterate at a menu of all things, lack of AA/AF etc options leave a sour taste in my mouth. At least Nvidia and ATI threw us a bone. Honestly after seeing some of the bugs first hand, and what I mentioned above I shudder to think about what Diablos gonna be like.

This was full screen and me looking at the afterburning reading for fps randomly on my g15, not very scientific I know! I'm sure someone will get into to the dirty and post something proper though.

All Ultra (which some are saying is bugged)
1920x1200
5870@875/1250
860@3.8 4gb ram

 

[Qbah]

I did some testing in STARCRAFT 2 with different AA modes and made graphs FPS vs Time(s). Common settings:

Rig:
Q9450 @ 3.2GHz
HD5850 @ 850 / 1150
4GB RAM
Win7 64bit HP
Cat 10.7a

Game:
Mission - Outbreak (running around with 40-50 Reapers killing stuff)
1920x1080
All Ultra
V-sync ON (via D3D Overrider)
Triple Buffering ON (via D3D Overrider)
Mouse Smoothing ON

No AA:

MIN - 17
MAX - 62
AVG - 40


MSAA x4 Box:

MIN - 5
MAX - 53
AVG - 31


SSAA x4 Box:

MIN - 5
MAX - 31
AVG - 21

All in all, NoAA and MSAA felt smooth and great. SSAA looked amazing, but you could feel the sluggishness compared to the less demanding modes. I think I'm gonna stick to MSAA for SC2.


EDIT: The minimum numbers are weird... I guess the drops were so short, they weren't drawn on the graphs?


EDIT2: Here's another graph, AAA x4 Box used. Different mission though ("Welcome to the jungle"). Definitely less action on this one compared to the earlier one. Still, game's running great

MIN - 0 (no idea when that happened )
MAX - 63
AVG - 38


EDIT3: Here's a Q6600 @ 3GHz with a stock HD4870 512MB, no AA, all else maxed. 4GB RAM Win7 Pro 64bit. 1920x1200. Mostly the night cycle, the end part is during day.

 

[Skurge]

I hope you aren't worried - half of that is more than enough for a RTS.

I remember seeing threads where people debated that more than a certain number of fps was actually prejudicial for RTS.

I'm not I'm actually happy, I consider 30 to be fine for an RTS. Then again, I won't be missing AA in an RTS.

 

[Xarick]

I would like to see those numbers with vsync and mouse smoothing off.

 

[Skurge]

I would like to see those numbers with vsync and mouse smoothing off.

How does mouse smoothing effect performance? I thought it just well smoothed out mouse movements.

 

[Sind]

Seems that playing in Windowed mode with just the taskbar visable, AA effectiveness is reduced, looks almost off, but not quite when compared with it off. Returns to normal 8xAA in Windowed (fullscreen) and fullscreen.

Just wanted to add for anyone messing around that control alt F adds in game fps counter. Happy trails!

 

[Sylvanas]

Thanks for the numbers Qbah. Is it just box mode that works with these driver workarounds in SC2? I have not seen any CSAA numbers or wide/narrow/edge detect filters used (I realize the wide/narrow filters are inferior to box anyway). Still looks to be quite playable with 4xAA + 5850 @ 1920 .

 

[Qbah]

AAA = serious glitches during cut scenes. Gonna stick with MSAA x4.

 

[lifeblood]

So is AA in SC2 "worth it" (knowing that "worth it" is subjective). I would not have thought an RTS would benefit from AA. Of course, other then Civ IV, I don't play RTS's. It just doesn't seem like something that would see any benefit as I didn't think that eye candy was the big draw, as opposed to first person shooters. I actually prefer Civ III over Civ IV because it doesn't have all that distracting eye candy.

AMD didn't plan on putting in AA in SC2 and I would have agreed with that. It was the same decision Bioware made. Obviously the outcry was loud enough they reconsidered their decision and put it there. I guess that's good as they are listening to thier customers.

 

[konakona]

In most games that I know of, in-game smooth mouse feature (or something along that line) did little outside adding some input lag. I wonder how well or poorly it works in SC2, regardless its impact on performance (probably negligible if any)

 

[Obsoleet]

I tend to stay away from beta drivers at all costs due to my experience with Nvidia non-WHQLs (and in recent years, their WHQLs). But I wanted to see what this looked like.

I was running 4x/16x with SC2 and it seems fine. When I tried 8x/16x the main menu was stuttering. Which from other people's accounts, is not really right. This is at 1920x1200, and dual monitor if that would make any difference (beta drivers running a seemingly poorly optimized game and dual monitor at that). I bumped it down to 4x/8x just for good measure.

 

[Cerb]

So is AA in SC2 "worth it" (knowing that "worth it" is subjective). I would not have thought an RTS would benefit from AA. Of course, other then Civ IV, I don't play RTS's. It just doesn't seem like something that would see any benefit as I didn't think that eye candy was the big draw, as opposed to first person shooters.

Moving jagged edges need AA, as do stationary jagged edges with very high contrast to the adjacent pixels, because many of us see the jagged edges more than the shapes they are supposed to look like. It's worth it in every 3D game. Some may need lower levels than others, but they all benefit from it. What we really need is to move towards setting values of pixels by blending single samples per poly weighted by actual display area taken up in that square of the screen, rather than just blending evenly spaced point-samples.

I actually prefer Civ III over Civ IV because it doesn't have all that distracting eye candy.

I prefer it because it's more fun, and doesn't have all that this-is-one-reason-why-Python-shouldn't-be-embedded-into-your-application lag. Civ III was made of awesome, and will be hard to top.